Network interface not UP after migration

[cloudcafetech]

I migrate Ubuntu VM from VMware to Kubevirt using forklift, migration went well but in kubevirt migrated VM network interface did not UP.

Source (Vmware) VM:

Destination (Kubevirt) VM:

Plan:

apiVersion: forklift.konveyor.io/v1beta1
kind: Plan
metadata:
  name: vm-mig-plan
  namespace: konveyor-forklift
spec:
  preserveStaticIPs: true
  archived: false
  description: ''
  map:
    network:
      name: vmware-network-map
      namespace: konveyor-forklift
    storage:
      name: vmware-storage-map
      namespace: konveyor-forklift
  provider:
    destination:
      name: host
      namespace: konveyor-forklift
    source:
      name: vsphere-provider
      namespace: konveyor-forklift
  targetNamespace: virtualmachines
  vms:
    - hooks: []
      name: vw-jumphost
  warm: false

[mnecas]

Hi, seeing that the device name changed after migration, the netplan did not bring up the interface.
We do have a script to inject udev rules and rename the interface to the original name but it's focused on the static IPs.
Think we could reuse it for this case.

[cloudcafetech]

@mnecas Thanks for pointing me interface change.

script to inject udev rules and rename the interface to the original name

Can you please share script which will inject udev rules and rename the interface to the original name and how to use.

[mnecas]

Sorry for the late response, you can find the script here https://github.com/kubev2v/forklift/blob/main/pkg/virt-v2v/customize/scripts/rhel/run/network_config_util.sh
We would need to run it even with DHCP I'll need to double-check if it would cause some problems.

[mnecas]

Ah after reading the code, we do it also for the DHCP but that script is missing the cloud.cfg so all we need to do is to update the script to look at the cloud init configs and create the udev rules from it.

Scratch that ^. Not sure if this is possible because we can't map the devices without the static IPs. We could add the cloud init config with static IP but not sure about DHCP.

As workaround you could try to inject a udev rules yourself before the migration starts this way it will preserve the interface name after migration.
https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/8/html/configuring_and_managing_networking/consistent-network-interface-device-naming_configuring-and-managing-networking#network-interface-naming-schemes_consistent-network-interface-device-naming

[cloudcafetech]

@mnecas Thanks.

Can you help me one more thing though its not V2V migration but its live migration inside cluster.

I am using Kubevirt along with Multus & whereabouts.

My network attachment definition as follows

apiVersion: "k8s.cni.cncf.io/v1"
kind: NetworkAttachmentDefinition
metadata:
  name: bridge-whereabouts
  namespace: virtualmachines
spec:
  config: '{
      "cniVersion": "0.3.1",
      "name": "bridge-whereabouts",
      "type": "bridge",
      "bridge": "br0",
      "promiscMode": true,       
      "ipam": {
        "type": "whereabouts",
        "range": "192.168.0.0/24",
        "range_start": "192.168.0.20",
        "range_end": "192.168.0.50",
        "gateway": "192.168.0.1",
        "routes": [
          { "dst": "0.0.0.0/0" }
        ]
      }
    }'

After successful migration all good except IP change from 192.168.0.20 to 192.168.0.21

I want keep same IP (192.168.0.20) when VM launch first time.